globally overloading new

[Paul Thomas]

John,

I have had a response from the gcc maintainers concerning the slowness of
new/delete with gcc-3.3 onwards.  The problem is indeed sjlj and bypassing
std::bad_alloc effects a very considerable speed-up to new.  I thought to
give it a whirl with octave by globally overloading new with the test
version they have sent me.  Do you have a suggestion as to where best this
should be inserted so as to effect the global overload from one place?

Paul T